I don't see where using substitution adds instability or has a problem with accuracy, unless they change the ASCII value for the character '@'.

It fixes one symptom but fails to address at least six serious bugs lurking in the hand-rolled parsing code, including the root cause of this problem.

Sometimes, getting all fancy for such a little straight forward script is not worth the effort.

Sometimes, shorter and simpler code is easier to maintain -- especially in this case, where this particular bug would not have been present at all. If that makes code "fancy" and that's a bad thing, so be it.